Intelligent Agents for Children's Early Math Skills: The Role of Scaffolds, Parent Attitudes, and Age

• In Event: 28.029 - Young Children's STEM Learning From Media: The Important Role of Parent Scaffolds and Confidence

Abstract

Introduction. In the 21st century, intelligent agents will be able to interact with children, responding contingently to what children say. This study examined the impact of an intelligent interface, either with a popular media character who spoke to children or a no character voiceover, on young children’s learning of the add-one rule, that adding one to a number increases the total by one unit (Baroody, Purpura, Eiland, & Reid, 2015). The game included scaffolds (see Vygotsky, 1978) to determine what aids children needed to complete the task successfully. We also examined children’s age and parent enjoyment of math as predictors of the number of rounds in which children needed scaffolds during game play. We then examined the role of scaffolds for transfer of learning of the add-one rule from the 2D game to a 3D context.

Method. Four-year-old children (n = 94; Mage = 4.89; 45 girls) played the intelligent math game using a Wizard of Oz procedure in which a hidden adult responded contingently to children’s verbal interactions through an onscreen interface. The game was comprised of 3 rounds with 4 add-one problems per round. The number of rounds (0-3) in which children needed scaffolds to correctly answer the add-one problems was recorded during data collection. Parents (N = 71; 51 mothers, 11 fathers, 6 “parents”) also answered paper and pencil surveys about their attitudes about math, as well as their child’s competency in math.

Results. OLS regression analyses were computed on the number of rounds with scaffolds needed to perform add‐one problems with age, parent enjoyment of math, and condition (character or no character) as predictors. The overall model was significant, F (3, 67) = 6.88, p < .001, adjusted R2 = 0.20. Older children (p = .001) and parents who reported enjoying math more when they were in school (p = .003) needed fewer rounds of scaffolds in the game. Condition was not significant. We then conducted an OLS regression analysis with transfer scores as the dependent variable with age, condition, and number of rounds in which scaffolds were needed as independent variables. The overall model was again significant, F (3, 91) = 8.76, p < .001, adjusted R2 = 0.20. Children who required fewer levels of scaffolds (p < .001) and were in the character condition (p = .02) were significantly more likely to answer transfer math problems correctly. Initial analyses indicated that parent enjoyment of math was not a significant predictor of transfer.

Conclusion. Math skills are an important 21st century skill for children’s future educational and occupational success. In the current study, we found that parent enjoyment of math while they were in school, as well as children’s age, predicted how much additional scaffolding was needed to master an early math concept. Mastery of the add-one concept and the character condition then predicted transfer scores, regardless of children’s age. These findings suggest that both parents and Intelligent Agents can contribute to the "zone of proximal development" (Vygotsky, 1978) in shaping young children's early math skills.

Authors

• Sandra Calvert, Georgetown University

• Naomi R. Aguiar, Georgetown University

• Marisa Putnam, Georgetown University

• Charlotte Wright, Georgetown University

• Angella Liu, Georgetown University

• Evan Barba, Georgetown University